ISTITUTO DI STUDI E ANALISI ECONOMICA
The contribution of domestic, regional, and
international factors to Latin America’s
business cycle
by
Melisso Boschi
University of Perugia, and
Centre for Applied Macroeconomic Analysis (CAMA) Via A. Pascoli, 20, 06123 – Perugia – Italy
e-mail: mboschi@stat.unipg.it and
Alessandro Girardi
University of Rome “Tor Vergata” and,
ISAE, Piazza dell'Indipendenza, 4, 00185 – Rome – Italy e-mail: a.girardi@isae.it
The Series “Documenti di Lavoro” of the Istituto di Studi e Analisi Economica – Institute for Studies and Economic Analyses (ISAE) hosts the preliminary results of the research projects carried out within ISAE. The diffusion of the papers is subject to the favourable opinion of an anonymous referee, whom we would like to thank. The opinions expressed are merely the Authors’ own and in no way involve the ISAE responsability.
The series is meant for experts and policy-makers with the aim of submitting proposals and raising suggestions and criticism.
La serie “Documenti di Lavoro” dell’Istituto di Studi e Analisi Economica ospita i risultati preliminari di ricerche predisposte all’interno dell’ISAE: La diffusione delle ricerche è autorizzata previo il parere favorevole di un anonimo esperto della materia che qui si ringrazia. Le opinioni espresse nei “Documenti di Lavoro” riflettono esclusivamente il pensiero degli autori e non impegnano la responsabilità dell’Ente.
La serie è destinata agli esperti e agli operatori di politica economica, al fine di formulare proposte e suscitare suggerimenti o critiche.
Stampato presso la sede dell’Istituto
ABSTRACT
This paper quantifies the relative contribution of domestic, regional and international factors to the fluctuation of domestic output in six key Latin American (LA) countries: Argentina, Bolivia, Brazil, Chile, Mexico and Peru. Using quarterly data over the period 1980:1-2003:4, a multivariate, multi-country time series model was estimated to study the economic interdependence among LA countries and, in addition, between each of them and the three world largest industrial economies: the US, the Euro Area and Japan. Falsifying a common suspicion, it is shown that the proportion of LA countries’ domestic output variability explained by industrial countries’ factors is modest. By contrast, domestic and regional factors account for the main share of output variability at all simulation horizons. The implications for the choice of the exchange rate regime are also discussed.
Keywords: International business cycle, Latin America, exchange rate regimes, Global VAR methodology, VEC models.
NON TECHNICAL SUMMARY
Over recent years, the increasing international economic integration driven by the liberalisation of current and capital accounts has stimulated a growing number of studies on the causative determinants of macroeconomic fluctuations in emerging markets. This work aims to analyse to what extent domestic, regional and international economic conditions affect domestic output fluctuations in six key Latin American (LA) countries - namely Argentina, Bolivia, Brazil, Chile, Mexico and Peru - and the implications for the choice of the exchange rate regime.
To quantify the relative contribution of domestic, regional and international shocks in explaining domestic output fluctuations, quarterly data over the period 1980:1-2003:4 was used and a multi-variate time series model was estimated to include those six LA countries as well as three major industrial economies (the US, Euro Area and Japan). The econometric methodology consists of a procedure for aggregating a number of VEC systems in a Global Vector Auto Regressive (GVAR) model describing the world economy in order to perform dynamic simulation exercises. Using quarterly data over the period 1980:1-2003:4, nine country/region-specific Vector Error Correction (VEC) models were estimated, each containing four endogenous domestic variables (output, real interest rate, real exchange rate, net foreign assets), two foreign variables (foreign output and foreign real interest rate) and the price of oil. Country-specific foreign variables, constructed as weighted averages of the endogenous variables of the other countries/regions, and the real oil price are modelled as weakly exogenous.
IL CONTRIBUTO DEI FATTORI NAZIONALI, REGIONALI ED
INTERNAZIONALI AL CICLO ECONOMICO DELL’AMERICA
LATINA
SINTESI
In questo lavoro si misura il contributo relativo dei fattori nazionali, regionali ed internazionali alle fluttuazioni del prodotto interno delle sei principali economie dell’America Latina: Argentina, Bolivia, Brasile, Cile, Messico e Perù. Utilizzando serie storiche trimestrali per il periodo 1980:1-2003:4, si presenta un modello multi-paese per analizzare le interdipendenze economiche reciproche tra questi paesi e l’influenza esercitata dalle tre più grandi economie mondiali: gli Stati Uniti, l’Area euro e il Giappone. In contrasto a quanto ottenuto in precedenti lavori, si mostra che la quota di variabilità del prodotto interno dei paesi latino-americani spiegata dalle economie industrializzate è modesta. Per contro, i fattori nazionali e regionali rappresentano la fonte principale delle fluttuazioni macroeconomiche in riferimento all’intero orizzonte di simulazione. Si discutono, inoltre, le implicazioni per la scelta del regime di tasso di cambio. Parole chiave: Ciclo economico internazionale, America Latina, regimi di tasso
CONTENTS
1 INTRODUCTION
...7
2 MODELLING LATIN AMERICAN ECONOMIES IN A
MULTICOUNTRY
FRAMEWORK
...9
2.1 The GVAR model
...10
2.2 Generalised Forecast Error Variance Decomposition
...12
3 PRELIMINARY
ANALYSES
AND ESTIMATION RESULTS
...14
4 ASSESSING
THE
GEOGRAPHICAL ORIGIN OF BUSINESS
CYCLE FLUCTUATIONS IN LATIN AMERICA
...23
5 CONCLUSIONS
...28
APPENDIX
...29
1 INTRODUCTION
1In keeping with the central message of the Optimal Currency Areas (OCAs) literature initiated by Mundell (1961) and McKinnon (1963), detecting the sources of business cycle has important implications for the choice of exchange rate regimes. If, in fact, one economy is hit by shocks dissimilar to those hitting its trading partner countries, the cost of adopting a fixed exchange rate regime, and thus giving up monetary policy, can be correspondingly large. The canonical criteria suggested by early contributions to OCAs (e.g. Artis (2003), HM Treasury (2003)) also state that if the standard pre-requisites for successful currency area hold, a fixed exchange rate regime may gain stability before adverse shocks make it fail. In many academic and policy circles, these criteria, although more than forty-years-old, are still considered to be a useful framework to consult when deciding upon the adoption of a common currency.
Following the currency and financial crises of the nineties, and especially the Argentine turmoil of 2001-2002, a wide debate has concerned the choice among available currency regimes options for Latin American countries (e.g. Edwards (2002), Berg et al. (2002)). This work aims to analyse to what extent domestic, regional and international economic conditions affect domestic output fluctuations in six key Latin American (LA) countries - namely Argentina, Bolivia, Brazil, Chile, Mexico and Peru - and the implications for the choice of the exchange rate regime. This country sample is chosen mainly to compare more easily our results to those of the existing literature to be reviewed below, and especially Ahmed (2003) and Canova (2005). Our analysis is naturally related to the strand of research studying the comovement of LA countries’ business cycles with each other and with developed economies’. Hoffmaister and Roldos (1997) document that domestic country-specific aggregate supply shocks are by far the most important source of output fluctuations in LA countries. Aiolfi et al. (2006) uncover a sizeable common component in LA countries’ business cycles using common dynamic factors techniques, thus suggesting the existence of a regional cycle. On the other hand, Agénor et al. (2000) point out that the business cycle in 12 developing countries is positively related to the output and real interest rate fluctuations in industrial economies, albeit they do not try to quantify the importance of external shocks compared to domestic ones. Employing a Bayesian dynamic latent factor model, Kose et al. (2003) and Kose et al. (2008) estimate the world, region and country-specific components in output, consumption and investment of sixty countries covering seven regions.
1
As far as concerns Latin America, Kose et al. (2003) find that country-specific factors explain the largest part of the variance of output in all LA countries considered in this study, with the exception of Bolivia, for which the regional world component is more important than the region and country-specific one.
From a wider perspective, our analysis is also related to the literature on the link between international business cycle and the choice of a proper exchange rate regime for a small open economy. Berg et al. (2002) find that supply shocks in LA countries are weakly correlated among them and, most importantly, with the US ones, providing evidence against the adoption of a common currency in the region or against straight “dollarisation”. Ahmed (2003) focuses on the existence of the prerequisites for six LA countries to adopt a fixed exchange rate regime with their main trading partners (the US). While domestic business cycle seems to be driven by US monetary policy rather than by foreign output shocks, external shocks taken as a whole (foreign output, US interest rates, terms of trade) explain a smaller component of LA business cycle than domestic shocks (output, real exchange rate, inflation); this results points towards the adoption of a freely floating exchange rate. By contrast, Canova (2005) finds that US monetary policy shocks, magnified by the interest rates transmission channel, are a relevant source of fluctuations of LA countries’ inflation and output.
The critical difference between the papers cited above and our study is three-fold. First, besides the US we also consider the Euro Area and Japan as possible sources of external shocks to domestic business cycle in LA countries. This is partly motivated by the trade relationship between LA and Euro Area countries. But, as it will become apparent below, this is not the entire story since financial linkages - through NFA and short-term interest rates - play a determinant role. Second, we examine the role exerted by neighbour countries on each LA country’s business cycle in order to assess the existence of the pre-requisites for the adoption of a common currency area. Third, our empirical framework is explicitly designed to identify shocks according to their geographical origin. The latter point is particularly important when comparing our results to those obtained by Kose et al. (2003) and Kose et al. (2008). In fact, while they can only recover the different components of the variables of interest, using the GVAR methodology it is possible to identify the role played by specific foreign economies to domestic business cycle.
interest rate, real exchange rate, net foreign assets), two foreign variables (foreign output and foreign real interest rate) and the price of oil. This is consistent with a parsimonious, reduced form, small open economy model such as that presented in Boschi (2007). Country-specific foreign variables, constructed as weighted averages of the endogenous variables of the other countries/regions, and the real oil price are modelled as weakly exogenous.
The main findings can be summarised as follows. First, domestic factors explain by far the largest share of domestic output variability over all simulation horizons in all LA countries. Second, regional factors, though much less important than domestic ones, contribute to the variability of domestic output more than industrial countries’ ones. This is true for all LA countries except Mexico. Third, in all LA countries the proportion of the forecast error variance of output explained by industrial countries factors is overall modest. These results should inform the choice between freely floating and fixed exchange rate regimes. Also, they should be taken into account when choosing a reference currency in a fixed exchange rate arrangement: “dollarisation” does not appear an obvious option. Aside from their scientific merits and policy implications, our findings that international risk sharing could be problematic at a regional level but it is still viable when capital crosses continents is consistent with the conclusions in Aiolfi et al. (2006) and may also be of benefit to international investors.
The remainder of the paper is structured as follows. Section 2 reviews the inter-regional macro-econometric framework. Section 3 presents preliminary analysis on the individual series as well as the main estimation results relative to country/region VEC systems and the properties of the GVAR model. The quantitative assessment of the geographical sources affecting output fluctuations in LA countries is discussed in Section 4 along with the main policy implications. Concluding remarks follow.
2
MODELLING LATIN AMERICAN ECONOMIES IN A
MULTI-COUNTRY FRAMEWORK
describe the steady-state configuration which the model tends to revert to in the long-run. The advantages of the GVAR over panel cointegration techniques are well-known (Baltagi (2004) and Pesaran et al. (2004b)) and relate to the possible distortion of within-group cointegration test results caused by the existence of between-group cointegration, as shown by Banerjee et al. (2004). Also, the GVAR allows for a coherent analysis of short-run dynamics of the systems through scenario simulations.
Specifically, the GVAR methodology consists of a procedure for stacking in a single coherent model of the world economy a number of country-specific VEC systems and explicitly allows for interdependences across economies in a true multi-country setting. The crucial advantage of this methodology is that although the shocks hitting the variables of the global system are unidentified according to their economic nature (for instance, supply, demand or policy disturbances), nevertheless they are identified basing on their geographic origin. This is because each country/region-specific system in the multi-country model is estimated conditionally on foreign variables, thus leaving only modest correlation among cross-country shocks to endogenous factors. Thus, our empirical framework makes it possible to distinguish and identify the shocks which originated in the three industrial countries/regions (US, Euro Area and Japan), in addition to those which originated in each LA country, rather than considering only one country (commonly the US in the previous literature) or an ambiguous “rest of the world” as the main source of external shocks.
2.1 The GVAR model
Adopting the same notation as in Pesaran et al. (2004a), there is benefit in reviewing the econometic setup employed in this work. There are N + 1 countries/regions in the world economy indexed by
i
= i =0,1,...,N 2. For each country the following VEC model is estimated3:
*
0 2 [ , 1 ( 1)] 0 0
it i i it i i i i t− i t i it i t it
Δ =x a +a D +Π κ −Π v −κ − +Λ Δ +x Ψ Δ +d ε (1) where xit is a (ki×1) vector of country
i
domestic variables,* it
x is a (ki*×1) vector of foreign variables specific to country i (to be defined below),
d
t is a(
k
d×
1)
vector of I(1) variables common to all country-specific models and2 N =8 in this paper. i=0 is the reference country (the US). 3 The exposition refers to a VARX*
exogenous to the global economy (such as oil prices),vi t,−1 ≡(z′i t,−1,d′ ′t−1) , *
( , ) it ≡ ′ ′ ′it it
z x x , ai0 is a (ki×1) vector of fixed intercepts, ai2 is a (ki×m) matrix of coefficients of the exogenous deterministic components included in the
(m×1) vector Dit, Λi0 is a
*
(ki×ki ) matrix of coefficients associated to the foreign variables, Ψi0 is a (ki×kd) vector associated to the global variables,
it
ε is a (ki×1) vector of country-specific shocks, with εit∼N 0 Σ( , ii), where Σii is a non-singular variance-covariance matrix, and where t =1, 2,...,T indexes time. The number of long-run relations is given by the rank
r
i≤
k
i of the*
( )
i i i d
k × k +k +k matrix Πi. Finally, in order to avoid introducing quadratic trends in the levels of the variables when Πi is rank-deficient, ki −ri restrictions
1
i = i i
a Π κ are imposed on the trend coefficients, where ai1 is the coefficient of the time trend term in the isomorphic level VAR form of (1) and κi is a
*
(ki +ki +kd) 1× vector of fixed constants.
The foreign variables x*itare weighted averages of the variables of the rest of the world with country/region-specific weights, wij, given by trade shares, i.e. the share of country jin the total trade of country
i
over the years 1995-2001, measured in 1995 US dollars. Thus a generic foreign variable* it
x is given by:
* 0 N it ij jt j x w x = =
∑
(2) where wii =0, ∀ =i 0,1,...,N and N0 ij 1 j= w =∑
, ∀i j, =0,1,...,N . In our set-up, all foreign variables collected in the vector x*itas well as the global exogenous variables, dt, are treated as long-run forcing variables.Rather than estimating directly the complete system composed by the 1
it = i t
z W x , where Wi is the
*
(ki +ki )×k matrix collecting the trade weights ij
w , ∀i j, =0,1,...,N .
Therefore, for each country/region the following VAR form of model (1) is obtained:
0 1 2 1 0 1 1
i i t = i + it+ i it + i i t− + i t + i t− + it
A W x a a a D B W x Ψ d Ψ d ε (3)
where Ai and Bi are matrices of dimension ki×(ki +ki*) and matrix Ai has full row rank. Stacking the N+1 systems (3) yields the following GVAR in level form:
0 1 2 1 0 1 1
t = + t+ t + t− + t + t− + t
Gx a a a D Hx Ψ d Ψ d ε (4)
where G is a
k k
×
full rank matrix, ah =(a0h,...,aNh)′ for h=0,1, 2, 0 0 ( ,..., N N)′ = G A W A W , H=(B W0 0,...,B WN N)′, for h=0,1, 0 ( ,..., ) h = h Nh ′Ψ Ψ Ψ , for h=0,1, Dt =(Ψ0t,...,ΨNt)′. The GVAR has the reduced form: 0 1 2 1 0 1 1 t = + t+ t +F t− + t+ t− + t x b b b D x ϒ d ϒd u (5) where 1 h h − = b G a for h=0,1, 2, = −1 F G H , 1 h h − = G Ψ ϒ for h=0,1, and 1 t t − = u G ε .4
4 As pointed out by Pesaran et al. (2004a), three conditions need to be fullfilled so as to ensure that the
GVAR estimation procedure is indeed equivalent to the simultaneous estimation of the VAR model of the world economy. First, the global model must be dynamically stable, i.e. the eigenvalues of matrix
F in equation (5) lie either on or inside the unit circle. Second, trade weights must be such small that
2
0 0
N ij j= w →
∑
as N→ ∞, ∀i. Third, the cross-dependence of the idiosyncratic shocks must be sufficiently small, so that 0 ,1 N
ij ls j
N∑= σ → ∞, ∀i l s, , , where σij ls, =cov(ε εilt, jst) is the covariance of
2.2 Generalised Forecast Error Variance Decomposition
The bulk of our empirical investigation is conducted using the Generalised Forecast Error Variance Decomposition (GFEVD) developed by Koop et al. (1996) and Pesaran and Shin (1998). The GFEVD considers the proportion of the variance of the n-step ahead forecast error of the variable of interest which is explained by conditioning on the non-orthogonalised shocks ujt, uj t, 1+ , ...,
, j t n
u + for j =1,...,k, while explicitly allowing for the contemporaneous correlations between these shocks and the shocks to the other equations in the system5. Although this methodology prevents a structural interpretation of the impulses, it overcomes the identification problem by providing a meaningful characterisation of the dynamic responses of variables of interest to typically observable shocks6. One useful feature of the GFEVD is its invariance to the ordering of the variables. Formally, the proportion of the n-step ahead forecast error variance of the lth element of xt accounted for by the innovations in the
th
j element of xt can be expressed as:
-1 1 2 0 ( ) ( ) 1 1 0 ( ) GFEVD( ; ; ) n n jj l l j l t j t n n n l l n − = − − = ′ σ = ′ ′
∑
∑
F F F s G Σs x u G ΣG s (6) 0,1, 2,... n= ; l =1,...,k; j =1,...,kwhere all symbols are defined above7.
5 It is worth emphasising that this is the reason why the GFEVD encompasses simpler methods
traditionally used to assess cross-country business cycle asymmetry such as the correlation analysis of shocks (e.g. Berg et al. (2002)).
6 We resort to GFEVD because it is impossible to recover the structural shocks from the GVAR residuals
due to the large number of variables whose contemporaneous relationship is ignored. In the GVAR estimated in this paper, including ki =4 endogenous variables for each of the N+ =1 9 country models, exact identification of shocks would require 108 (i.e.
∑
iN=0k ki( i−1)) restrictions derived byeconomic theory, which seems an impossible task to undertake. Dees et al. (2007a) identify the shocks to US monetary policy by imposing a recursive structure on the US block of the variance-covariance matrix of the GVAR. However, this exercise is beyond the scope of this paper.
7 Notice that due to the possible non-diagonal form of matrix Σ, the elements of GFEVD across j
3 PRELIMINARY
ANALYSES
AND ESTIMATION RESULTS
Data description. Time series data for the following countries/regions were
considered: Argentina, Bolivia, Chile, Brazil, Mexico, Peru, the US, Japan and the Euro Area. We use quarterly seasonally adjusted series covering the period 1980:1-2003:48. The Euro Area variables were constructed as weighted averages of the corresponding time series of the following countries in the region, with weights given by the per capita PPP-GDP share of the period 1995-2000: Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, and Spain9. For each country/region, a VEC model (1) was estimated, where the vector of endogenous variables, xit, includes
y
t, srt,t
q and
nfa
t, denoting real per-capita output, short-term real interest rate, real exchange rate and the net foreign asset/nominal GDP ratio respectively; the vector of country-specific foreign variables, x*it, includes* t
y
and srt*, representing the rest of the world real per-capita output and short-term interest rate, respectively; finally, the vectord
t includes the oil price in real terms, oilt, as a global weakly exogenous variable10. The matrix of trade weights used to construct the country/region-specific foreign variables is reported in Table 1, where the 1995 - 2001 trade shares are displayed in column by country/region. The Appendix indicates in detail data sources and variables construction.Tab. 1 Trade weights
Source: Direction of Trade Statistics Yearbook, IMF, 2002.
Notes: Trade weights, computed as shares of exports and imports in 1995-2001, are displayed in column by country/region. Each column, but not row, sums to one.
8 Note that the 1980s mark the beginning of the modern wave of international capital flows to Latin
America and thus analysing the role of this factor in domestic business cycle prior to the sample start makes little sense.
9 On the validity of the aggregating expedient to construct synthetic time-series for the Euro Area
economy as a whole see Girardi and Paesani (2008) among others.
10 Boschi (2007) motivates the inclusion of these variables in the GVAR basing on a small open economy
Unit root tests. As a preliminary exercise, we carried out standard ADF unit
root tests on the time series involved. Panel [A] of Table 2 reports results based on AIC order selection, while statistics shown in Panel [B] use the modified AIC method proposed by Ng and Perron (2001) to correct the size distortion ofordinary ADF test statistics.
Tab. 2 ADF unit root test statistics
Notes: The ADF statistics are based on univariate AR(p) models in the levels with p chosen according to the modified AIC, with a maximum lag order of 11. The sample period is 1980:1-2003:4. The regressions for all variables in the levels include an intercept and a linear trend with the exception of interest rates whose underlying regressions include only an intercept. The 95 percent critical value for regressions with trend is -3.46 and for regressions without trend -2.89.
(2002, 2003). The results are reported in Table 3, Panels [A] and [B]. Since the distribution under the null hypothesis is non-standard, we use the critical values provided by Lanne et al. (2002).
Tab. 3 ADF unit root tests with breaks statistics
Overall, the combination of both types of tests (standard and with breaks), indicate that all variables can be reasonably considered to be driven by I(1) stochastic trends. On the other hand, differencing the series appears to induce stationarity11.
Determination of the autoregressive order. We chose the lag length of the
endogenous variables,
p
i, by combining standard selection criteria; namely the Akaike information criterion (AIC), the Schwarz Bayesian criterion (SBC) and the log-likelihood ratio statistic (LR). These criteria were adjusted to take into account the potential small sample problems, starting from a maximum lag order of four. The results, reported in Table 4, indicate that the SBC suggests order one for all models except Bolivia, Mexico and US, the AIC selects order four for Chile, Mexico and the Euro Area, order three for Peru, order two for Argentina, Bolivia, Japan and the US, and order one for Brazil, while the LR favours an order of autoregression higher than four for Mexico, three for Chile, Peru, and Euro Area, two for Bolivia, Japan, and US, one for Argentina and Brazil.Given the alternatives, and taking into account the limited sample size compared to the number of unknown parameters in each VARX* model, where X* indicates foreign exogenous variables, the lag order
p
i is set equal to 1. This choice is comforted by the fact that the SBC estimates the lag order consistently, while the AIC does not (Lütkepohl (2006), p. 151). In order to choose the lag order of the foreign specific variables, qi, an unrestricted VAR was run for each country/region in which the foreign variables are treated as endogenous, obtaining similar results12. Basing on this evidence and considering data limitations, we set qi equal to one in all models.Misspecification tests. The selected lag order and the inclusion of dummy
variables corresponding to residual values larger than 3.5 times the standard error is sufficient to obtain a satisfactory specification of the models, giving
11 The only exceptions are the real exchange rate of Mexico that seems to be stationary, and the net
foreign assets of Bolivia, which appear to be I(2). We choose to model these variables as realizations of I(1) processes since the actual integration properties of the real exchange rate series of Mexico are likely to depend on the composition of its trading partners prices and exchange rates. For example, using a different basket of trading partners, Boschi (2007) finds that the real exchange rate of Mexico isI(1). The net foreign assets of Bolivia were treated as I(1) since this hypothesis is rejected at the 5 percent confidence level but not at the 10 percent.
Tab. 4 Test statistics for selecting the lag order of the endogenous (domestic) variables in the VARX*(pi,qi) model
support to our model specification strategy. Univariate specification tests, reported in Table 5, show that the null hypothesis of no serial correlation is rejected only in 5 out of 36 equations at the standard confidence level, while the null of normality is rejected only in 3 equations. Finally, the univariate F test rejects the null of homoschedasticity only for Japanese output and US real exchange rate at 5 percent level.
Tab. 5 Univariate specification tests statistics
In order to detect possible parameters instability due to structural breaks conventional CUSUM and CUSUMSQ tests at single equation level for each model were undertaken. The results, unreported here to preserve space, were comforting since episodes of parameters instability emerge only for a limited number of equations and only for very short periods of time13.
Cointegration tests. Table 6 reports the maximum eigenvalue and trace tests
statistics together with their associated 90 and 95 percent critical values. Both tests select unambiguously a cointegration rank equal to 1 for Brazil, Mexico,
Tab. 6 Cointegration rank statistics
Notes: the last two columns report the critical values at the 95 percent and 90 percent significance level. Statistics in bold indicate acceptance of the null hypothesis at the 5 percent significance level.
13 These are the beginning of the nineties for the Argentinean, Chilean, Peruvian, and US net foreign
Peru, and Japan, and 4 for the US. For the other models, where the results were less clear cut, we favoured the conclusion of the trace test comforted by Johansen (1992), according to which the maximum eigenvalue test may produce a non-coherent testing strategy. Thus, we set a cointegration rank of 1 for Argentina, and 2 for Bolivia and the Euro Area. As for Chile, after considerable experimentation, a rank of 2 was chosen in order to have a more stable Global VAR14.
Properties of the Global VAR. Since in the GVAR the total number of
endogenous variables is 36 and that of cointegrating relations is at most 1515, it then follows that matrix F in equation (5) must have at least 36-15=21 eigenvalues that fall on the unit circle in order to ensure stability of the global model. Our results confirm this; the matrix F estimated from the country-specific models has exactly 21 eigenvalues falling on the unit circle, while the remaining 15 are all less than one (in modulus).
A second key assumption of the GVAR approach is that idiosyncratic shocks are cross-sectionally weakly correlated. The basic idea is that conditioning the estimation of country/region-specific VEC models on foreign variables considered as proxies of “common” global factors will leave only a modest degree of correlation of the remaining shocks across countries/regions. This is also important if we were to interpret the disturbances in the GFEVD analysis as “geographically structural”: an external shock is truly external if its contemporaneous correlation with internal shocks is weak. In order to verify these claims, contemporaneous correlations of residuals across different country-specific models for each equation were computed. Table 7 reports such correlation coefficients, computed as averages of the correlation coefficients between the residuals of each equation (variable) with all other countries/regions equations residuals. A two-tailed t-test rejects the hypothesis that these coefficients are significantly different from zero at the conventional level. Thus, the model seems to be successful in capturing the effect of common factors driving domestic variables.
A third econometric concern refers to the assumption that foreign variables and oil price are weakly exogenous in the country/region-specific VEC models. Along the lines described by Johansen (1992) and followed by Pesaran et al.
14 Notice that the long-run structure defined by the cointegration space of each country/region specific
model could be restricted according to the implications of a small open economy model (e.g. Boschi (2007) and Dees et al. (2007b)), but given the explicit focus of this paper on the relationship among economies at a business cycle frequency, we limited our exercise to unrestricted models.
15 That is the sum of the ranks of matrix
i
Tab. 7 Average cross-section correlations of residuals
Notes: each entry is the average correlation of the residual of the equation on the corresponding row for the country/region on the corresponding column with all other countries/regions endogenous variables residuals. Two-tailed t-test statistics with 93 d.o.f. are in square brackets. The null hypothesis is no correlation. The 5 percent critical value is 1.98.
(2004a), we examined the weak exogeneity of these variables by testing the joint significance of the error correction terms in auxiliary equations of the country/region-specific foreign variables, x*it and the oil price. Specifically, we carried out the following regression for each lth element of country i vector of foreign variables, x*it and for the oil price:
* , , 1 , 1 , 1 i r j il t il ijl i t i t il t j x ECM − − = ′ = μ +
∑
γ + ϕ v + ζ Δ Δ where μil is a constant, , 1 j i tECM − , j =1,...,ri, are the estimated error correction terms corresponding to the ri cointegrating relations found in the
th
i model, ,
il k
ϕ are coefficients, Δvi t, 1− is defined by (1), and ζil t, is the residual. Then an
F test of the joint hypotheses that γ =ijl 0, j=1,...,ri, is carried out. Table 8 reports the results.
Tab. 8 F statistics for testing the weak exogeneity of the
country-specific foreign variables and oil prices
Most of the test statistics are not significant at the 5 percent level16. Given the overall statistical support and the strong theoretical prior in favour of the weak exogeneity hypothesis, foreign variables and the oil price were treated as weakly exogenous.
4 ASSESSING
THE
GEOGRAPHICAL ORIGIN OF BUSINESS
CYCLE FLUCTUATIONS IN LATIN AMERICA
As discussed above, the modest degree of cross-country correlation of reduced form residuals allows for an approximated identification of disturbances according to their geographical origin. Given the focus of the present study, we confined our analysis to output fluctuations. Table 9 reports the GFEVD of each LA country’s domestic output over a simulation horizon of 40 quarters. Panel [A] refers to the contribution to domestic output forecast error variance of domestic shocks, i. e. y,
sr
, q, and nfa. Panel [B] summarises the contribution of external shocks classified according to whether their origin is regional, i.e. from other LA countries, or from one of the three industrial economies we consider in the analysis. Finally, Panel [C] reports an overall comparison of domestic versus foreign contribution to each country’s domestic output fluctuations.Domestic shocks. A mixed picture of the local determinants of output
variability emerged. Real factors (output itself) are neatly predominant over the whole forecast horizon only in Argentina and, especially, Brazil, while this is true only up to the 12th quarter for Bolivia, Chile and Mexico, and up to the 20th quarter for Peru. Financial factors seem to play a significant role in all countries apart from Argentina and Brazil (and even here still play a role)17. This is consistent with Canova’s (2005) findings that financial factors are an important channel of transmission of foreign shocks; or it could be interpreted as idiosyncratic sources of variability. However, this first block of results should be taken with caution since, as detailed above, the GFEVD tool does not allow for
16 The weak exogeneity assumption is rejected at the 1 percent level only in the model of Peru for the
short-term rates and in the Euro Area model for oil prices, while it is rejected at the 5 percent level in the models of Mexico and US for output.
17 Specifically, net foreign assets are the main source of variability in Chile (from the second simulation
Tab. 9 Generalized variance decomposition of the forecast error of output
Notes: share of the k-step ahead forecast error variance of domestic output explained by the shocks on the corresponding column. Entries have been normalized so that they sum to 100. Each entry in columns “All domestic factors” and “All foreign factors” are the sum of the corresponding percentages in columns 2, 3, 4, 5 and in columns 6, 7, 8, 9, respectively.
Regional vs domestic shocks. Over the entire forecast horizon, regional
factors contribute approximately 20 percent of domestic output variability in Argentina, Bolivia and Chile but drops to approximately 10 percent in Brazil and Mexico. This pattern is somehow more variegated in Peru where the contribution of regional shocks ranges from 13 to 42 percent. Overall this result supports evidence of a sizeable regional business cycle component in Latin America. Aiolfi et al. (2006) attribute this feature to the role of common global factors on the grounds of limited trade and financial linkages among these economies. However, the breakdown (unreported) of the figures in column 5 of Table 9 show that regional factors affect domestic business cycle through financial channels (short-term rates and net foreign assets) in a non-negligible way. Thus, since the main common global real and financial factors were controlled for in this study in a coherent model of the world economy, the findings are interpreted as due to similarities in the economic structure of the LA countries examined.
Industrial countries’ vs regional and domestic shocks. In all Latin American
countries considered here, domestic factors contribute far more than industrial countries’ factors to the variability of domestic output18. Overall, industrial countries explain a small fraction of output fluctuation, ranging from 7 percent in Bolivia to almost 13 percent in Mexico. Specifically, the US economy is the most important contributor to domestic output forecast variability at all horizons for Argentina and Peru. The role of Euro Area is never very large on impact, but tends to increase over time. Japan gives an important contribution to output variability in all countries, and especially in Argentina, Bolivia, Brazil and Chile. This central finding disputes the other relevant literature on international business cycles, most of which concentrate on the role of US macroeconomic variables and implicitly assume that the US role in the global economy and its trade and financial links with Latin America (the US “backyard”) are the main driving force behind business cycles co-movements in this region (Ahmed (2003), Canova (2005)). Falsifying a common suspicion, estimates show that the proportion of LA countries’ domestic output variability explained by the US (and by the other industrial countries) is modest when compared to the contribution of regional shocks.
Robustness checks. In order to gain some insights on the reasons why our
results differ from those studies where the US role seems bigger, a number of
18 This is true for all countries at all horizons, with an average difference between the percentage
alternative models were estimated19. In particular, we estimated first a VEC model including only output of all countries/regions considered in the GVAR - i.e. Argentina, Bolivia, Brazil, Chile, Mexico, Peru, the US, Euro Area and Japan. The results show that the role of the US and regional shocks are larger than in the GVAR, especially at longer forecast horizons, with the exception of Mexico for which the importance of US shocks decreases over time. In addition, six VEC models, one for each LA country — each model including the relevant LA country’s factors, i.e.
y
t , srt, qt and nfat, along with the US counterparts — were estimated. As expected, in these six models the US factors play an even bigger role than in the VEC model containing only output of all countries/regions. The US explain on average more than 20 percent of domestic output forecast error variance in all LA countries, with the only exception of Brazil.All in all, considering the evidence provided by the simple VEC models, the reason why in the GVAR the influence exerted by the US is smaller seems to be related more to the inclusion of a larger set of countries/regions than to the larger number of factors. This helps to understand why previous literature - where the US is the only external economy taken into account - overestimated the contribution of the US shocks to LA business cycle. In this respect, the paper by Kose et al. (2003) goes along the right direction since it considers a large group of countries. They find, like in this study, that country-specific factors are the main determinant of output fluctuations in Latin America, but they reserve a smaller role to the regional factors compared to this paper. However, the methodology in their paper, namely a Bayesian dynamic latent factor model, does not allow to recover the geographical origin of factors affecting domestic business cycle, but rather identifies the generic components of a series as divided in world, region and country-specific20. For this reason the GVAR appears a more suitable methodology to address the problem of choosing the proper exchange rate regime for an emerging market basing on the main geographical determinants of its business cycle.
Which exchange rate regime for Latin American countries? The findindgs of
this paper have important implications for the choice among such alternative extreme exchange rate regimes, i.e. hard pegs (currency board or unilateral
19 Results of these additional estimations are unreported to save space, but they can be provided by the
authors upon request.
20 Notice that from a more technical perspective, the methodology used in Kose et al. (2003) differs from
“dollarisation”), the formation of an independent common currency area and the freely floating exchange rate. First, as long as “dollarisation” requires a large degree of business cycle synchronisation among the country adopting the dollar and the US economy, the GFEVD analysis shows that in the LA countries this regime may be subject to strong destabilising shocks originated in countries other than the US, either developed or developing. A sensible way to take into account this fact could be pegging the domestic currency to a “synthetic” foreign currency built as a weighted average of the currencies of the main industrial and developing countries affecting domestic business cycle. Second, although the contribution of regional factors to domestic business cycle in LA countries is noticeable, and indeed larger than industrial countries influence, nevertheless idyosincratic shocks play a dominant role in all LA countries’ economies. This result cast doubts on the viability of a common currency area along the path set by the European Monetary Union. Idiosyncratic shocks could destabilise such a monetary arrangement well before it could enhance the required real and financial integration necessary to make it work. All results above suggest that a freely floating exchange rate might be the most viable option to be pursued in LA countries, in line with what argued by Ahmed (2003) and Berg et al. (2002).
Implications for portfolio diversification. Aside from the academic and policy
5 CONCLUSIONS
Over recent years, the increasing international economic integration driven by the liberalisation of current and capital accounts has stimulated a growing number of studies on the causative determinants of macroeconomic fluctuations in emerging markets. The vast majority of existing contributions implicitly assume that US are the main origin country of external shocks. In this paper we have demonstrated that this is not the case, at least not in LA countries.
To quantify the relative contribution of domestic, regional and international shocks in explaining domestic output fluctuations, quarterly data over the period 1980:1-2003:4 was used and a multi-variate time series model was estimated to include six key LA countries (Argentina, Bolivia, Brazil, Chile, Mexico and Peru) as well as three major industrial economies (the US, Euro Area and Japan). The main findings can be summarised as follows. Domestic and regional factors account for the main share of output variability at all horizons, while the proportion explained by industrial countries factors is modest. All in all, assessing the relevant contribution of shocks originating in other neighbour countries and in countries/regions other than the US will provide a better understanding of the actual geographical origin of external drivers of output variability in LA countries.
APPENDIX
A.1 Data sources
Net Foreign Assets (NFA). The NFA series is obtained for each country as the sum, period-by-period, of foreign assets and liabilities given by the following quarterly time series taken from the IFS database: DIA (Direct Investment Abroad - code 78...BDDZF), PIA(Portfolio Investment Assets - code 78...BFDZF), OIA (Other Investment Assets - code 78...BHDZF), DIL (Direct Investment Liabilities - code 78...BEDZF), PIL(Portfolio Investment Liabilities - code 78...BGDZF), and OIL (Other Investment Liabilities - code 78...BIDZF). Therefore: NFA=DIA+PIA OIA DIL+ − −PIL OIL− .
Population (POP). The source is the IFS database. The code is 99Z..ZF.... Available annual data are interpolated linearly.
Nominal Output (YNC). The series is the volume of GDP in billions of national currency. It is taken from IFS for all countries except for Brazil. The code is 99B./CZF.... The series for Brazil is obtained from IPEADATA.
Output (YCC). The source for all countries, except Brazil, is the IFS database. The code is ..99BVP/RZF.. (2000=100). The quarterly data for Argentina’s GDP volume index are only available from 1993:1; the series is extended backward using the rates of growth of the GDP index series provided by Oxford Economic Forecasting. The GDP index of Brazil is obtained by deflating (with the CPI) the GDP volume in billions of national currency provided by IPEADATA.
Price index (CPI). The source is the IFS’ Consumer Prices Index (CPI), which code is 64...ZF.. (2000=100).
Exchange rates (NER). The source is the IFS’ series of National Currency per US Dollar, with code 17 .RF.ZF... except fo Mexico for which the series ..WF.ZF... is used.
Nominal short-term interest rates (SR). The series is the Money Market Rate or equivalent (code 60B..ZF...) from the IFS.
Oil price (OILP). The series is the price of Brent from IFS, with code 11276AAZZF....
A.2 Variables construction
Development Indicators 2002. Following Pesaran et al. (2004a), the variables used in the estimation of each country/region-specific VEC model are constructed from the series above as follows:
2000
ln[100 ( / ) / ]
y= ⋅ YCC POP POP ; 1
0.25 ln(1 /100) ln( / )
sr = ⋅ +SR − CPI+ CPI ; 2000
ln(100 / ) ln( )
q= ⋅NER NER − CPI ;
/( / )
nfa=NFA YNC NER ; * 0 N i j ij j y =
∑
= w y ; * 0 N i j ij j sr =∑
= w sr ; 2000 ln(100 / )REFERENCES
Agenor, P. R., C. J. McDermott, and E. S. Prasad (2000) “Macroeconomic Fluctuations in Developing Countries: Some Stylized Facts” World Bank Economic Review, 14, 251-285.
Ahmed, S. (2003) “Sources of Economic Fluctuations in Latin America and Implications for Choice of Exchange Rate Regime”, Journal of Development Economics, 72, 181-202.
Aiolfi, M., L. Catão, and A. Timmermann (2006) “Common Factors in Latin America’s Business Cycles”, IMF Working Paper, 06/49.
Artis, M. J. (2003) “Reflections on the Optimal Currency Area (OCA) Criteria in the Light of EMU”, International Journal of Finance and Economics, 8, 297-307.
Baltagi, B. H. (2004) “Comment”, Journal of Business Economics and Statistics, 2, 163-164.
Banerjee, A., M. Marcellino, and C. Osbat (2004) “Some Cautions on the Use of Panel Methods for Integrated Series of Macro-economic Data”, Econometrics Journal, 7, 322-340.
Berg, A., E. Borensztein, and P. Mauro (2002) “An Evaluation of Monetary Policy Regime Options for Latin America”, North American Journal of Economics and Finance, 13, 213-235.
Boschi, M. (2007) “Foreign Capital in Latin America: A Long-run Structural Global VAR Perspective”, University of Essex, Department of Economics Discussion Paper, n. 647.
Canova, F. (2005) “The Transmission of US Shocks to Latin America”, Journal of Applied Econometrics, 20, 229-251.
Dees, S., F. di Mauro, M. H. Pesaran, and L. V. Smith (2007a) “Exploring the International Linkages of the Euro Area: A Global VAR Analysis”, Journal of Applied Econometrics, 22, 1-38.
Dees, S., S. Holly, M. H. Pesaran, and L. V. Smith (2007b) “Long-run Macroeconomic Relations in the Global Economy”, Economics - The Open-Access, Open-Assessment E-Journal, 2007-3.
Edwards, S. (2002) “The Great Exchange Rate Debate after Argentina”, North American Journal of Economics and Finance, 13, 237-252.
Girardi, A. and P. Paesani (2008) “The Transfer Problem in the Euro Area: A Cointegration Analysis”, Open Economies Review, forthcoming.
HM Treasury (2003) “Analysis of European and UK Business Cycles and Shocks”, available at www.hmtrasury.gov.uk.
Johansen, S. (1992) “Determination of the Cointegration Rank in the Presence of a Linear Trend”, Oxford Bullettin of Economics and Statistics, 54, 383-397.
Koop, G., M. H. Pesaran, and S. M. Potter (1996) “Impulse Response Analysis in Nonlinear Multivariate Models”, Journal of Econometrics, 74, 119-147. Kose, M. A., C. Otrok, and E. Prasad (2008) “Global Business Cycles:
Convergence or Decoupling?”, IZA Discussion Papers, n. 3442.
Kose, M. A., C. Otrok, and C. Whiteman (2003) “International Business Cycles: World, Region, and Country Specific Factors”, American Economic Review, 93, 1216-1239.
Lane, P. R. and G. M. Milesi-Ferretti (2004) “The Transfer Problem Revisited: Net Foreign Assets and Real Exchange Rates”, Review of Economics and Statistics, 86, 841-857.
Lanne, M., H. Lütkepohl, and P. Saikkonen (2002) “Comparison of Unit Root Tests for Time Series with Level Shifts”, Journal of Time Series Analysis, 23, 667-685.
Lanne, M., H. Lütkepohl, and P. Saikkonen (2003) “Test Procedures for Unit Roots in Time Series with Level Shifts at Unknown Time”, Oxford Bulletin of Economics and Statistics, 65, 91-155.
Lütkepohl, H. (2006) New Introduction to Multiple Time Series Analysis, Springer Verlag, Berlin.
McKinnon, R. I. (1963) “Optimum Currency Areas”, American Economic Review, 53, 717-725.
Mundell, R. A. (1961) “A Theory of Optimum Currency Areas”, American Economic Review, 51, 657-665.
Ng, S. and P. Perron (2001) “Lag Lenght Selection and the Construction of Unit Root Tests with Good Size and Power”, Econometrica, 69, 1519-1554. Pesaran, M. H., T. Schuermann, and S. M. Weiner (2004a) “Modelling Regional
Interdependencies using a Global Error-Correcting Macroeconometric Model”, Journal of Business Economics and Statistics, 2, 126-162.
Pesaran, M. H., T. Schuermann, and S. M. Weiner (2004b) “Rejoinder to Comments on Modelling Regional Interdependencies using a Global Error-Correcting Macro-econometric Model”, Journal of Business Economics and Statistics, 2, 175-181.
Pesaran, M. H. and Y. Shin (1998) “Generalised Impulse Response Analysis in Linear Multivariate Models”, Economics Letters, 58, 17-29.
Working Papers available:
n. 47/05 M. MALGARINI
P. MARGANI B.M. MARTELLI
Re-engineering the ISAE manufacturing survey
n. 48/05 R. BASILE
A. GIUNTA
Things change. Foreign market penetration and firms’ behaviour in industrial districts: an empirical analysis
n. 49/05 C. CICCONI Building smooth indicators nearly free of
end-of-sample revisions
n. 50/05 T. CESARONI
M. MALGARINI G. ROCCHETTI
L’inchiesta ISAE sugli investimenti delle imprese manifatturiere ed estrattive: aspetti metodologici e risultati
n. 51/05 G. ARBIA
G. PIRAS
Convergence in per-capita GDP across European regions using panel data models extended to spatial autocorrelation effects
n. 52/05 L. DE BENEDICTIS
R. DE SANTIS C. VICARELLI
Hub-and-Spoke or else? Free trade agreements in the “enlarged” European Union
n. 53/05 R. BASILE
M. COSTANTINI S. DESTEFANIS
Unit root and cointegration tests for cross-sectionally correlated panels.
Estimating regional production functions
n. 54/05 C. DE LUCIA
M. MEACCI
Does job security matter for consumption? An analysis on Italian microdata
n. 55/05 G. ARBIA
R. BASILE G. PIRAS
Using Spatial Panel Data in Modelling Regional Growth and Convergence
n. 56/05 E. D’ELIA Using the results of qualitative surveys in
quantitative analysis
n. 57/05 D. ANTONUCCI
A. GIRARDI
Structural changes and deviations from the PPP within the Euro Area
n. 58/05 M. MALGARINI
P. MARGANI
Psychology, consumer sentiment and household expenditures: a disaggregated analysis
n. 59/05 P. MARGANI
R. RICCIUTI
Working Papers available:
n. 60/05 M. BOSCHI
A. GIRARDI
Euro Area inflation: long-run determinants and short-run dynamics
n. 61/05 M. BOVI Book-Tax Gap. An Income Horse Race
n. 62/06 M. BOVI The Cyclical Behavior of Shadow and Regular
Employment
n. 63/06 G. BRUNO
C. LUPI
C. PAPPALARDO G. PIRAS
The cross-country effects of EU holidays on domestic GDP’s
n. 64/06 M. COZZOLINO
F. DI NICOLA M. RAITANO
Il futuro dei fondi pensione: opportunità e scelte sulla destinazione del TFR
n. 65/06 S. LEPROUX
M. MALGARINI
Clima di fiducia e spesa delle famiglie in Italia: un’analisi disaggregata secondo il reddito degli intervistati
n. 66/06 M. BOVI Consumers Sentiment and Cognitive
Macroeconometrics Paradoxes and Explanations
n. 67/06 G. ROCCHETTI Modelli di business nel mercato del software e
partecipazione delle imprese italiane al fenomeno open source
n. 68/06 L. CROSILLA La stagionalità delle inchieste ISAE su imprese
e consumatori: aspetti metodologici ed evidenza empirica
n. 69/06 C. OLDANI Money demand & futures
n. 70/06 R. BASILE
S. DE NARDIS A. GIRARDI
Pricing to market of italian exporting firms
n. 71/06 B.M. MARTELLI
G. ROCCHETTII
The ISAE Market Services Survey:
Methodological Upgrading, Survey Reliability
n. 72/06 M. FIORAMANTI Predicting sovereign debt crises using artificial
neural networks: a comparative approach
n. 73/06 S. ZECCHINI
M. VENTURA
Working Papers available:
n. 74/06 G. FERRI
M. VENTURA
Macchinari del made in Italy e dinamiche dei distretti industriali
n. 75/07 R. BASILE Intra-distribution dynamics of regional
per-capita income in Europe: evidence from alternative conditional density estimators
n. 76/07 M. BOVI National Accounts, Fiscal Rules and Fiscal
Policy Mind the Hidden Gaps
n. 77/07 L. CROSILLA
S. LEPROUX
Leading indicators on construction and retail trade sectors based on ISAE survey data
n. 78/07 R. CERQUETI
M. COSTANTINI
Non parametric Fractional Cointegration Analysis
n. 79/07 R. DE SANTIS
C. VICARELLI
The “deeper” and the “wider” EU strategies of trade integration
n. 80/07 S. de NARDIS
R. DE SANTIS C. VICARELLI
The Euro’s Effects on Trade in a Dynamic Setting
n. 81/07 M. BOVI
R. DELL’ANNO
The Changing Nature of the OECD Shadow Economy
n. 82/07 C. DE LUCIA Did the FED Inflate a Housing Price Bubble? A
Cointegration Analysis between the 1980s and the 1990s
n. 83/07 T. CESARONI Inspecting the cyclical properties of the Italian
Manufacturing Business survey data
n. 84/07 M. MALGARINI Inventories and business cycle volatility:
an analysis based on ISAE survey data
n. 85/07 D. MARCHESI The Rule Incentives that Rule Civil Justice
n. 86/07 M. COSTANTINI
S. de NARDIS
Estimates of Structural Changes in the Wage Equation: Some Evidence for Italy
n. 87/07 R. BASILE
M. MANTUANO
La concentrazione geografica dell’industria in Italia: 1971-2001
n. 88/07 S. de NARDIS
R. DE SANTIS C. VICARELLI
The single currency’s effects on Eurozone sectoral trade: winners and losers?
n. 89/07 B.M. MARTELLI
G. ROCCHETTII
Working Papers available:
n. 90/08 M. MALGARINI Quantitative inflation perceptions and
expectations of Italian Consumers
n. 91/08 P. L. SCANDIZZO
M. VENTURA
Contingent valuation of natural resources: a case study for Sicily
n. 92/08 F. FULLONE
B.M. MARTELLI
Re-thinking the ISAE Consumer Survey Processing Procedure
n. 93/08 M. BOVI
P. CLAEYS
Treasury v dodgers. A tale of fiscal consolidation and tax evasion
n. 94/08 R. DI BIASE Aliquote di imposta sul lavoro dipendente:
analisi per figure tipo e con dati campionari
n. 95/08 M. BOVI The “Psycho-analysis” of Common People’s
Forecast Errors. Evidence from European Consumer Surveys
n. 96/08 F. BUSATO
A. GIRARDI A. ARGENTIERO
Technology and non-technology shocks in a two-sector economy
n. 97/08 A. GIRARDI The Informational Content of Trades on the
EuroMTS Platform
n. 98/08 G. BRUNO Forecasting Using Functional Coefficients
Autoregressive Models
n. 99/08 A. MAJOCCHI
A. ZATTI
Land Use, Congestion and Urban Management
n. 100/08 A. MAJOCCHI Theories of Fiscal Federalism and the
European Experience n. 101/08 S. de NARDIS
C. PAPPALARDO C. VICARELLI
The Euro adoption’s impact on extensive and intensive margins of trade:
the Italian case n. 102/08 A. GIRARDI
P.PAESANI
Structural Reforms and Fiscal Discipline in Europe
n. 103/08 S. TENAGLIA M. VENTURA
Valuing environmental patents legal protection when data is not availableby
n. 104/08 P. L. SCANDIZZO M. VENTURA
